Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

Transcription

1 Two Problems in Textual Interpretation Author(s): Umberto Eco Source: Poetics Today, Vol. 2, No. 1a, Roman Jakobson: Language and Poetry (Autumn, 1980), pp Published by: Duke University Press Stable URL: Accessed: 31/10/ :01 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org. Duke University Press is collaborating with JSTOR to digitize, preserve and extend access to Poetics Today.

2 TWO PROBLEMS IN TEXTUAL INTERPRETATION UMBERTO ECO Semiotics, Bologna & Literature, Yale PREFACE The two sections below are two further elaborations of the theory of the interpretation of narrative texts presented in The Role of the Reader (1979a) and then developed further in Lector in Fabula (1979b). For the convenience of the reader, I am presenting here the table of levels of textual cooperation published in Eco 1979a, p. 14. In the box on discursive structures I did not sufficiently develop the voice "chosen of the isotopies," since the concept of isotopy was there understood as used in Greimas' semiotics. As to the deeper intensional levels, in The Role of the Reader I have developed only a few aspects of the question since my major interest was in the interpretation of the narrative level (fabula) and in the extensional inferences (= possible worlds). In the two following sections I shall deal explicitly with isotopies and with some problems concerning the intensional interpretation of deep textural structures. 1. ON ISOTOPY. In Eco (1979a: 0.6.3) I devoted several pages to the notion of topic, defined as a cooperative device activated by the reader (usually in the form of a question) for the purpose of identifying the isotopy for interpreting the text. I wrote: "the topic as question is an abductive schema that helps the reader to decide which semantic properties have to be actualized, whereas isotopies are the actual textual verification of that tentative hypothesis," by which I meant to say that the topic as such is not expressed by the text, while the isotopy is a verifiable semantic property of it. In other words, the topic is a pragmatic device, while the isotopy is a level of possible semantic actualization of the text. Yet in order to analyze that semantic property or indeed that level of meaning that a text manifests, it is necessary to specify more exactly what is meant by isotopy. My hypothesis is that the term, variously defined by Greimas and by his school, is an umbrella term, a rather general notion that can allow for various more specific ones defining different textual phenomena. Only the clarification of these differences will make it possible to throw light on the positive theoretical aspects of the notion.? Poetics Today, Vol. 2:la (1980),

3 146 UMBERTO ECO INTENSIONSEXTENSIONS L Figure 1 9. ELEMENTARY '10 WORLD STRUCTURI ES IDEOLOGICAL STRUCTURES World matrices I I Assignment of truth va] lues 8. ACTANTIAL STRUCTURES Judgements of Actantial roles as manifested accessibility among woi rids by actorial roles Recognition of l~i I propositional attitudes 6. NARRATIVE STRUCTURES Macro-propositions i of the fabula 7. FORECASTS AND (themes, motives, INFERENTIAL WALK narrative functions) - Fabula as temporal I I succession or world stat :es 4. DISCURSIVE STRUCTURES Probability disjunction Individuation and inferences of topics Reduction Semantic of frames ] disclosures 5. (BRACKETED) Blowing up EXTENSIONS and narcotizing Isotopies - First uncommitted r references properties to a (possible) world. ACTUALIZED CONTENT 1. CODES AND SUBCODES Basic dictionary Rules of coreference Contextual and circumstantial selections Rhetorical and stylistic overcoding Common frames Intertextual frames Ideological overcoding 2. CIRCUMSTANCES OF UTTERANCE Informations about the sender, time and social context of the message, suppositions about the nature of the speech act, etc. Greimas (1970: 188) defines isotopy as "a complex of manifold semantic categories making possible the uniform reading of a story." The category would then have the function of textual or transsentential disambiguation, but on various occasions Greimas furnishes examples dealing with sentences and outright noun phrases. For instance, in order to explain in what sense the amalgam on a single classeme (either semantic category or repeated contextual seme) makes possible a uniform reading, he gives the example of the two expressions "le chien aboye"

4 TWO PROBLEMS IN TEXTUAL INTERPRETATION 147 and "le commissaire aboye." Given that "bark" has two classemes, human and canine, it is the presence of the dog or the commissioner that reiterates one of the two that decides whether "bark" is taken in a literal or figurative sense. It should be obvious that what are called classemes here are our contextual selections (cf. Eco, 1976: ). The human presence of the commissioner introduces a "human" context and makes it possible to make the appropriate selection out of the compositional spectrum of "bark."' But can we say that an isotopy obtains always and only under such conditions? Aside from the fact that if so, it would not differ from normal semantic coherence or from the notion of amalgamation, the lists made of the various meanings of the term either in Greimas or his disciples (cf. Kerbrat-Orecchioni, 1976) do say that at various times there are isotopies that are semantic, phonetic, prosodic, stylistic, enunciative, rhetorical, presuppositional, syntactic or narrative. It is therefore fair to assume that isotopy has become an umbrella term covering diverse semiotic phenomena generically definable as coherence at the various textual levels. But is that coherence obtained at the various textual levels by applying the same rules? That is why it is advisable, if not to work out an isotopic system, to make the term less equivocal and more manageable, at least for the purposes of this present paper, stipulating the minimum conditions for its use. In an initial examination, the meanings shown in Figure 2 seem to emerge. The diagram is not meant to finalize an isotopic system, but to show how the category can assume various forms. Figure 2 wsenith paradigmatic disjunction discursive ^ isotopies ^' sentential between sentences -with syntagmatic disjunction with paradigmatic disjunction with syntagmatic disjunction narrative isotopies (or at deeper levels) connected with discursive isotopic disjunctions unconnected with discursive isotopic disjunctions exclusive complementary Let us now consider a few examples that will support this classification. 'Cf. Greimas (1966:52-53). Cf. also van Dijk "Aspects d'une theorie generative du texte poetique" in Greimas ed., Essais de semiotiquepoetique, Paris: Larousse, 1972, pp , "It can be said that the central isotopy of a text is made up of the lowest seme or classeme dominating the greatest number of lexemes of the text."

5 148 UMBERTO ECO 1.1 Sentential discursive isotopies with paradigmatic disjunction.2 In his essay on crossword writing, Greimas (1970) examines this definition with the correlated denomination (1) I'ami des simples = herbalist in which the clever definition arises from the fact that simples has two contextual selections, one common and one specialized, precisely governed by the "vegetable" selection. Only after it is decided (through topicalization) that the term is understood in the second sense is it established that it counts as a substantive and not an adjective, and therefore it is decided to decode ami as "lover" or "fan" and not as "friend." The topic has intervened as a reading hypothesis (speaking of plants and not of ethical attitudes), and has pointed toward the appropriate contextual selection, and has imposed a rule of interpretive coherence affecting all the lexemes involved. We can apply the term isotopy to the semantic result of that coherent interpretation, and recognize the actualized isotopy as the "objective" content of the expression (objective in the sense that it is supported by the code. Naturally, in the case of this expression, which is deliberately ambiguous, or, if we like, bi-isotopic, the objective contents are two, both actualizable). It should be noted too that in this case the isotopy does not depend on any redundancy of a semantic type, since ami and simples do not seem to have semes in common. In truth the final isotopy is realized by the whole syntagm "question + solution": "the herbalist is the friend of simples." That is to say that once the topicalization is made (that the subject is plants) we get the sentence "The herbalist likes herbs" in which "botanist" imposes a vegetal seme which makes it possible to actualize the appropriate contextual selection within the componential spectrum of "simples." Cases of the same kind are demonstrated in those puzzles called mnemonic "cryptographs" studied extensively by Manetti and Violi (1977). That is the reason these isotopies are defined as sentential although at first glance they seem to apply only to defined descriptions. In each case they are characterized by paradigmatic disjunction: they depend on the fact that the code includes lexical expressions with a multiple meaning. It is evident that the paradigmatic disjunction derives from a cotextual pressure that operates syntagmatically, but that does not eliminate the need to decide what reading passage to assign to one or more componential spectrums. Moreover, these isotopies are denotatively exclusive: the subject is either the evangelical in spirit, or it is herbs. The topic intervenes as a concurrent, cooperative hypothesis to individuate contextual selections Sentential discursive isotopies with syntagmatic disjunction. Transformational grammar has accustomed us to ambiguous sentences like 2The distinction between isotopies with paradigmatic disjunction and those with syntagmatic disjunction corresponds to the one between vertical and horizontal isotopies proposed by Rastier and discussed in Kerbrat-Orecchioni, 1976:24-25.

6 TWO PROBLEMS IN TEXTUAL INTERPRETATION 149 (2) They are flying planes. which can generate two different deep structures. In disambiguating the sentence, paradigmatic disjunctions undoubtedly apply (it is necessary, for instance, to decide whether the verb should be understood as active or passive) but the fundamental decision (always deriving from the prior topicalization) is whether the subject is humans doing something with the airplanes, or airplanes doing something. At that point it is necessary to actuate a coreference and establish to whom or what they refers. We could say that the coreferential (syntagmatic) decision decides the paradigmatic choice concerning the meaning of the verb. These isotopies too are denotatively exclusive: either the subject is a human action, or it is mechanical objects. Here the topic intervenes as a concurrent hypothesis to actualize both coreferences and contextual selections Discursive isotopies between sentences with paradigmatic disjunction. Let us examine in this connection the little story of two fellows conversing during a party, cited by Greimas (1966). The first praises the food, the service, the hospitality, the beauty of the women, and finally the excellence of the "toilettes." The second replies that he has not yet been there. The second speaker, in interpreting the text uttered by the first, blunders because he superimposes two frames (cf. Eco, 1979). The frame "party" undoubtedly includes the hosts' garments, but cannot include the condition of the sanitary facilities, or it would have to consider also the electrical system, the water supply, the solidity of the walls, and the layout of the rooms. These elements are considered at most as belonging, say, to a frame like "interior architecture and furnishing." The party refers to a frame that is social in nature, furnishings to one that is technological. The individuation of the topic in this case is the individuation of the semantic field, so as to enable contextual selections to be effected. The French term "toilettes" is undoubtedly polysemic and acquires two meanings according to the disjunction between the selection "fashion" (which in turn belongs to a seme of "social" nature) and the selection "architecture." In this case we can certainly speak of the presence of a classeme or a dominant semantic category, since the text of the first speaker in fact abounded in key terms containing references to the party and to the social nature of the situation. There were misunderstandings possible, and the story makes us laugh precisely because it constitutes a case of awkward textual cooperation. These isotopies have paradigmatic disjunction because, if only on the basis of cotextual (syntagmatic) pressure, they concern contextual selections in lexemes with multiple meaning. These isotopes too are denotatively exclusive: the subject is either clothing or it is bathrooms. The topic intervenes as a concurrent cooperative hypothesis, to individuate contextual selections that hypothesize frames Discursive isotopies between sentences with syntagmatic disjunction. This is the case of the ambiguous sentence

7 150 UMBERTO ECO (3) Charles makes love with his wife twice a week. So does John. The point is whether this short text should be read as the story of two couples, or the story of a triangle. In this case too we have discursive isotopies with alternative denotations. In extensional terms, it is a matter of deciding whether there are three people involved or four. In order to do so, it is necessary to decide how "so" should be interpreted, but then it is a matter of establishing a coreference. The choice concerns the syntactic structure of the sentence, and only through a syntactic decision is the one or the other semantic result obtained. As already seen, it is through the topicalization operation that the decision is made as to whether the subject is two couples or a triangle: in the first case the logical structure of the text would be A:B=C:D while in the second it would be A:B = B:C. It is a problem of interpretive coherence; if four individuals are concerned and A and B are compared in the first sentence, "so" means that in the second sentence C and D should be compared; if, on the other hand, three people are involved and A and B are compared in the first sentence, "so" means that in the second B and C should be compared. But it is not obvious how the two interpretive decisions derive from the redundance of semantic categories. Here the connection is between the topic and coreferential decisions, without the mediation of contextual selections. At the most, as already seen, presuppositions of frames are involved. The two isotopies are characterized by syntagmatic disjunction. They are mutually exclusive (the subject is either the Kinsey report, or it is the story of adultery) but they are by no means denotatively alternative: some of the individuals remain the same in each case, only they are ascribed different actions and intentions. The topic intervenes as a cooperative concurrent hypothesis to establish the coreferences, thus orienting the structurization of different narrative worlds. 1.5 Narrative isotopes connected with isotope discursive disjunctions generating mutually exclusive stories. The following text is the French translation of an extract from Macchiavelli and it is irrelevant whether the original Italian text shows the same ambiguity as the French;3 the French text will be examined as if it were an anonymous original: (4) Domitian surveillait: l'age des senateurs, et tous ceux qu'il voyait en position favorable pour lui succeder il les abattait. II voulut ainsi abattre Nerva qui devait lui succeder. II se trouva qu'un calculateur de ses amis l'en dissuada, va que lui meme (my italics) etait arrive a un age trop avance pour que sa mort ne ffit toute proche; et c'est ainse que Nerva put lui succeder. It is immediately evident that here we have first of all a choice between two 3The text was proposed by Alain Cohen in the course of a colloquium on modalities held at Urbino at the International Semiotics Center in July Cohen's analysis, however, aimed at different goals from ours, and concerned only the discourse on Power referred to below.

8 TWO PROBLEMS IN TEXTUAL INTERPRETATION 151 transsentential discursive isotopies with syntagmatic disjunction: lui-meme can refer to either Domitian or Nerva. If it refers to Domitian, the death referred to later (sa mort) is the imminent death of Domitian; otherwise it is Nerva's death. It is therefore necessary to decide on the coreference on the basis of topicalization: is the subject Domitian's age or Nerva's? Once the coreference is decided, there is a denotatively alternative discursive sequence in respect to the other. In effect in one case the advisor tells Domitian not to kill Nerva because he - Domitian - will soon die and it is therefore useless to eliminate his possible successors; in the other, the advisor tells Domitian that Nerva will probably die soon and therefore does not present a danger for Domitian. But it is clear that two different stories can be derived on the basis of the two discursive isotopies. The two discursive isotopies generate two possible narrative recapitulations. In one case it is the story of a friend of Domitian's giving him an argument about Power. "In dying you risk losing Power, but by sparing Nerva and implicitly designating him your successor, though dying you retain control of the Power, you generate the new Power." In the other case it is the story of a friend of Nerva's making Domitian the victim of a courtier's wiles: "O Domitian, why do you want to kill Nerva? He's so old that he'll soon die by himself!" - and thus the courtier puts Nerva on the throne. Thus, two mutually exclusive stories emerge, whose individuation depends on the discursive actualization. Not only that, but at a deeper level (cf. Figure 1), there emerge different actantial structures and different ideological structures. According to Greimas' categories, the advisor can be seen as the Opponent of Domitian and Helper of Nerva, or as the Helper of Power and the Opponent of Domitian as a mortal, or a Helper of Domitian and neutral in regard to Nerva. And it can be decided that what is defined here is an ideological opposition of Power versus Death (in which Power overcomes even Death) or Power versus Shrewdness, where the courtier's wiles overcome the brutality of Power. It can also legitimately be asked whether it is the choice of coreferences that generates the different deep structures, or a preliminary hypothesis regarding the deep structures that in suggesting a specific topic controls the actualization of the coreferences at the discursive level. The interpretive cooperation is made of leaps and short-circuits at the different textual levels where it is impossible to establish logically ordered sequences. In each case we have seen that here the narrative isotopies are connected to these discourses (or vice versa). The two narrative isotopies are mutually exclusive but they are not at all denotatively alternative: in both cases the narration is about Nerva and Domitian, except that different actions and intentions are attributed to them. The individuals remain extensionally the same, but change some of their intensional features. Different possible worlds are developed. The topic intervenes to orient the structurization of these narrative worlds. 1.6 Narrative isotopies connected with isotopic discursive disjunctions that generate complementary stories.

9 152 UMBERTO ECO That is the case of the medieval theory of the four senses of the scriptural verses, also cited by Dante. Given the text (5) In exitu Israel de Aegypto - domus Jacob de papolo barbaro - facta est Judea santificatio ejus - Israel potestas ejus. We know that "if we look only at the letter it means the exodus of the children of Israel from Egypt in Moses' time; if we look at the allegory it means our redemption through Christ; if we look at the moral sense it means the conversion of the soul from the struggle and misery of sin to the state of grace; and if we look at the mystical sense it means the departure of the holy spirit from the servitude of this corruption to the freedom of eternal glory." Let us consider, in order to simplify matters, just the literal and moral senses. Once again everything depends on the topic hypothesis: Is the statement about Israel or about the human soul? The decision on this affects the discursive actualization: in the first case Israel will be understood as a proper name of a people, and Egypt as a proper name of an African country; in the second case Israel will be the human spirit, but then by interpretive coherence, Egypt will have to be sin (the reading levels cannot be mixed). Here, however, alternative senses of a componential spectrum will not be chosen, because we must foresee that in as rich an encyclopedia as the medieval one Israel denoted the chosen people and connoted the soul. Thus it is not like the case of "toilette" that has the sense of either x or y; here the expression connotes the sense y precisely because it denotes x. The relationship and implication is not one of disjunction. Consequently isotopic disjunction exists, but it is not based on disjunction but rather on semantic implication. Once the preferred reading at the discursive level is decided, various stories can be inferred from the actualized discursive structures, the the moral story will derive from the moral discursive actualization just as the literal one will from the literal discursive actualization. But the two stories (and we know in reality there are four) are not mutually exclusive; they are rather complementary, in the sense that the text can be read simultaneously in two or more ways, and one way reinforces the other rather than eliminates it. Thus narrative isotopies are connected with discursive isotopies, but are not mutually exclusive. On the contrary, they are denotatively alternative; the subject is either the chosen people, or it is the soul (and in fact the option is between various denotations and connotations). By virtue of this choice various possible worlds developed. The topic (both the discursive and the narrative) intervenes to choose between denotative and connotative semes, and orientate the structure of the narrative worlds Narrative isotopies not connected with discursive isotopic disjunctions that generate complementary stories in each case. Greimas (1970) in his analysis of the Bororo myth of the aras [a Brazilian name of a bird] speaks of another type of narrative isotopy. The myth in effect contains two stories, one about the search for water, and the

10 TWO PROBLEMS IN TEXTUAL INTERPRETATION 153 other about the problem of diet. So it is the "natural" isotopy versus the "alimentary." But in both cases we perceive that whatever the story (or thefabula) that we actualize, there is no change on the discursive level. The stories always tell of certain people and certain events. At the most, according to the narrative isotopy, we select certain actions as more pertinent than others, but the actions and people doing them remain the same, even if there is a change in the value we attribute to them in the narrative arrangement. It is a matter of elaborating a hypothesis with a narrative theme and relying on key terms or sentences without, however, paradigmatic disjunctions as to the sense of the lexemes, or syntagmatic disjunctions as to the sense of the coreferences. The persistence of a single discursive coherence results in this case in the two narrative isotopies not annulling each other, the relation between them not being exclusive or alternative, but complementary. Although Greimas chooses the alimentary isotopy as best, this does not mean that the story cannot be read through the natural isotopy as well. In fact the two isotopies reinforce each other. The toilet story is characterized by two opposing readings, of which one is clearly inferior, and if the first speaker had really wanted to speak of bathrooms, his utterance would have been conversationally inept because it violated the rule of relevance. That cannot be said about the myth of the aras. Thus we have here narrative isotopies unconnected with discursive disjunctions. The two or more narrative isotopies are not mutually exclusive. They are not even denotatively alternative; at most, different necessary features are attributed to different individuals (cf. Eco, 1979a: 8.7). Therefore, different possible narrative worlds are proffered. The topic intervenes only to orient the evaluation of the narratively pertinent features, and thus the structurization of those worlds Provisional conclusions. According to what has been said, it is permissible to assert that isotopy is an umbrella term covering various phenomena. Like all umbrella terms (such as iconism, presupposition, code) this one shows that the diversity conceals some unity. Indeed, isotopy refers almost always to constancy in going in a direction that a text exhibits when submitted to rules of interpretive coherence, even if the rules of coherence change according to whether what is wanted is to individuate discursive or narrative isotopies, to disambiguate definite descriptions or sentences and produce coreferences, to decide what things certain individuals do, or to establish how many different stories the same deed by the same individuals can generate. What should be clear in any case is that the identification of the topic is a cooperative (pragmatic) movement guiding the reader to individuate the isotopies as semantic features of a text. 2. USE AND INTERPRETATION An isotopy is a semantic feature of a text even when the text is open to several possible interpretations. We have seen that a given interpretation can choose one

11 154 UMBERTO ECO of two or more alternative isotopies or allow complementary isotopies to live together. It appears difficult to assert, on the other hand, that isotopies which a text does not allow can be attributed to it. This obviously negates Valery's statement that "il n'y a pas de vrai sens d'un texte." But that does not mean to say that there is just one and only one interpretation of a text. Between the theory that the interpretation is wholly determined by the author's intention and the theory that it is wholly determined by the will of the interpreter there is undoubtedly a third way. Interpretive cooperation is an act in the course of which the reader of a text, through successive abductive inferences, proposes topics, ways of reading, and hypotheses of coherence, on the basis of suitable encyclopedic competence; but this interpretive initiative of his is, in a way, determined by the nature of the text. By the "nature" of the text I mean what an interpreter can actualize on the basis of a given Linear Manifestation, having recourse to the encyclopedic competence toward which the text itself orients its Model Reader (cf. Eco, 1979a). However, we must distinguish between the free use of a text taken as imaginative stimulus and the interpretation of an "open" text. It is on that dividing line that the possibility of what Barthes calls a text of "jouissance" is grounded, without theoretical ambiguity: the decision must be taken as to whether the text is used as a text of "jouissance" or whether a certain text considers the stimulation of the freest possible use a constituent of its own strategy (and thus of its interpretation). But we believe that some restrictions obtain, and that the notion of interpretation always involves a dialectic between the strategy of the author and the response of the Model Reader. Naturally, aside from procedure, it is also possible to provide an esthetic of the free, deviant, desirous and malicious use of texts. Borges suggested reading the Odyssy as if it came after the Aeneid, or the Imitation of Christ as if it were written by Celine. These are splendid, exciting and realizable proposals, very creative indeed, because in effect a new text is produced (thus Pierre Menard's Quijote is quite different from Cervantes' which it accidentally corresponds to word for word). And then in the writing of that other text, it happens that the original is criticized, or hidden possibilities and values discovered in it. (Nothing is more revealing than a caricature because it resembles, but is not, the caricatured object, and on the other hand certain retold romances become more beautiful because they become "other" romances.) From the point of view of a general semiotics and in the light of the complexity of the pragmatic processes and contradictory nature of the Global Semantic Field, all these operations are theoretically explainable. But if the range of interpretations can be infinite, as Peirce has shown, the universe of discourse itself operates to limit the size of the encyclopedia. And an "other" text is nothing more than the strategy constituting the universe of these among its interpretations which, if not legitimate, are legitimizable. Any other decision on using a text freely amounts to a decision to expand the universe of discourse. The dynamics of unlimited semiosis (Eco, 1976) does not prevent it, but rather encourages it. But it is necessary to know whether what is wanted is to have the semiosis function, or to interpret a text.

12 TWO PROBLEMS IN TEXTUAL INTERPRETATION 155 Proust could read a railroad timetable and find in the names of the Valois region sweet, labyrinthian echoes of the Nervalian voyage in search of Sylvie. But it was not a matter of the interpretation of the timetable but rather of an almost psychedelic use of it. As a matter of fact, a railway timetable postulates only a specific type of Model Reader, able to deal with Cartesian orthogonal axes (vertical and horizontal) and with a vigilant sense of the irreversibility of temporal sequences. It seems, nevertheless, in the light of the table reproduced at the beginning of this article, that there are levels of actualization in which the reader seems more bound, with respect to a certain encyclopedic competence than in others. The reader who actualizes discursive structures is undoubtedly restricted by the nature of the lexicon (if the text says "a dog appeared" it cannot be made to say "a cat appeared"); similarly, it seems that on the level of narrative structures he has to recognize the worlds of the fabula as the text organizes them (if "there was once a woods with a little house where a little girl lived," it cannot be read "there was once a mountain with a castle where a king lived"). But what happens when, after the discursive and narrative structures are actualized, the readers advance hypotheses on deeper structures (actantial structures, ideological values)? Once the narrative structures are actualized, while the reader makes forecasts about the states of the fabula (delineating possible worlds), he can (before, after, or at the same time) formulate a series of macropropositions even more abstract than the narrative ones. That is, he can individuate the actorial roles (Greimas) or narrative sequences. He can deprive the actorial roles of their residual individuality and reduce them to actantial oppositions (subject/object; helper/ opponent; sender/addressee) deciding that in some cases a single actantial role is covered by more actors. What makes it difficult to define the theoretical placement of this cooperative mode is that, on the one hand, the reader has already to have precalculated hypotheses regarding the actants in order to be able to define certain narrative structures, and on the other hand, already to have delineated possible worlds with their individuals, in order to be able to determine the actors in the game. Take, for instance, a text like Gerard de Nerval's Sylvie. Each of the three women in it - Sylvie, Aurelia and Adrienne - enters into a game of ever-changing opposition with one of the others, donning various actantial roles, with one or the other becoming the real presence, as opposed to memory, according to the state of the fabula and the time section (present, immediate past, remote past) of which the narrator is speaking. Thus, on the one hand, the reader should already have proffered a hypothesis on the role of the character in that part of the fabula so as to formulate narrative macropropositions. On the other hand he should have recognized the states of the fabula in their logical sequence so as to determine whether a given discursive passage represents a fact that is happening, that happened, that is remembered, that in the past was believed and then was contradicted by subsequent reality, and thus lives. Obviously, possible worlds cannot be identified without the discursive structures having been actualized; but at the level of these latter, to disambiguate certain complexities of verb tenses, it would

13 156 UMBERTO ECO be necessary to have already formulated hypotheses not only about the worlds, but also about the actantial frame and the roles played by the characters. When we read Hugo's Quatre-vingt-treize at what point do we decide, on the explicit and repeated declarations of the author, that it is the story of a grandiose subject, Revolution, as the voice of the people and of God, that is projected against its exact opposite, Reaction? When is it fully understood than Lantenac or Cimourdain, Gauvain or the Convention, Robespierre or the Vandee are superficial manifestations of a deeper conflict about which and of which the author is chiefly speaking? And when does it happen that when this is understood, the reader foregoes individuating the characters, some "historical" and some fictitious, that populate the novel beyond the limits of the memorizable? It is clear that in an operation of that sort, the actantial hypothesis emerges not in resolution of a series of successive abstractions, from discursive structures to the fabula and from it to ideological structures, but rather establishes itself very quickly in the course of the reading, and guides the choices and forecasts, and determines macropropositional filterings. An action or an event can be left to fall, while the author's long philosophical perorations come to be part of what really pertains to the fabula, because across a multitude of aspects, gestures, and events we need to retain only what tells us what the revolution is doing to pursue its design, and how it works on the individuals and directs their actions. The same can be said for what we have called ideological structures, to which so much space is devoted in the textual research of the past decade. While an actantial framework preexists as an encyclopedic apparatus (even before it is realized in the text) like a system of oppositions, that is as an S-code (Eco, 1976), an ideological structure (whether at the level of encyclopedic competence or in its textual actualization) presents itself as a code in the true sense and that is as a system or correlations. We could thus say that an ideological structure manifests itself when axiological connotations are associated with actantial roles inscribed in the text. And it is when an actantial framework is provided with value judgements and the roles convey axiological oppositions like Good versus Evil, True versus False (or even Life versus Death or Nature versus Culture) that the text exhibits its ideology as if by a watermark. At that point the Model Reader's ideological competence comes in to direct the choice of the actantial framework and of the great ideological oppositions. For example, a reader whose ideological competence consists of a coarse but effective opposition between Spiritual Values (with the connotation of "good") and Material Values (with the connotation of "evil") might be tempted in a tale like Death in Venice to actualize two large oppositions on the actantial level; Aschenbach's esthetic vocation against his carnal desire (and thus Spirit versus Matter), assigning on the level of ideological structures a mark of "positive" to the first and "negative" to the second. Rather poor and hardly problematic reading, but an example was sought of how ideological competence determines the actualization of the deep textual structures. Naturally, a text can anticipate such competence in the true Model Reader and work - at all the appropriate

14 TWO PROBLEMS IN TEXTUAL INTERPRETATION 157 lower levels - to bring it to a head and induce the reader to individuate more complex ideological and actantial structures. Then there are the cases of deviant (more or less happy) decoding: a typical one is that of the Mysteres de Paris in which the ideological propensity of the proletarian readers functioned like a code shifter and induced them to actualize in revolutionary key a discourse made in a socio-democratic one. Ideological competence does not necessarily act as a restraint on the interpretation but may function as a stimulus as well. And, as such, it is conducive to finding in the text what the author was unaware of but which the text in some way conveyed.4 But what happens when the reader, while identifying the deep structures, brings to light something that the author could not have wanted to say but that the text nevertheless seems to show with absolute clarity? Obviously, here we are touching on the subtlest limit separating interpretive and hermeneutic cooperation. Furthermore, is it not proper to assume that it is the hermeneutic that discovers in the text the truth that it offers, discloses, allows to appear? Obviously there are hermeneutics and hermeneutics. The etymologies of Isidor of Seville - and many of Heidegger's - make words say what they cannot say if the encyclopedia has its own objective social existence; and the medieval readings of Virgil treated as a prophetic text did violence to the Virgilian discourse. These are the cases in which a text is not interpreted but rather used in absolute freedom, as if it were a pack of Tarots. But the case is different when a person peruses a text to draw conclusions from it on the deep impulses of the author or to find traces in it of his unavowed ideology. Sue wanted to be a revolutionary and wrote a book that was blandly reformist. But his working-class readers found calls to revolution in it. Who was right? Poe wished to tell the story of an extremely lucid mind, Dupin, and in the Dupin trilogy many have found the setting of a theatre of the unconscious. Is it right to disregard the author's numerous explicit assertions of Dupin's lucid and controlled rationality? Let us suppose there is a narrative text produced in recent years in which there are manifested obsessively - not only on the level of individuals, properties and relations but also on the level of the syntactic structures themselves - actantial impressions, anaphoric substitutions, abrupt switches from first to third person, in a word, difficulties in recognizing and making recognizable the subjects put into play by the enonce, and the same subject-author understood as enunciative strategy. It is not hard to apply this description to a long series of experimental or avant-garde texts. In such cases it may confidently be assumed that the author had in mind all those aspects of the current encyclopedia by which such expressive phenomena are related to dissociative elements and identity crises. The text is invested with and expresses as its proper content a schizomorphic vision - not described but manifested as style, as the organizational modality of the discourse. The author as empirical subject of the enonce may be more or less conscious of 4Cf. e.g. our studies on James Bond, Les mysteres de Paris, Superman, etc. in Eco, 1965a, 1965b, 1968, 1976.

15 158 UMBERTO ECO what he was doing, but textually he did do it, in the same way in which I may not know that a particular word has a particular meaning, but if I pronounce it, I've said what I've said. Just as on the psychological level one would speak of a "slip," it could be said that I spoke in a state of mental cloudiness, that I am stupid, that I made a mistake. At this point we come, however, to a different situation. Let us illustrate it with another text, from an epoch in which so many psychiatric and psychoanalytical discoveries were not yet in the public domain (or by a contemporary author but with a very limited encyclopedia). The text tells an ostensibly irrelevant story, but gives the clear impression that through the use of obsessive metaphors or the particular syntactic arrangements, the picture of a schizoid attitude or Oedipus complex is being constructed as if in a watermark. Could we say that this structure is part of the contents of the text that the Model Reader is called upon to actualize? By interpretation is meant the semantic actualization of what the text as strategy wants to say with the cooperation of its own Model Reader - in the ways and on the levels outlined above. It could then be maintained that a text that through its own structures manifests the schizoid personality of the author, or the fact that he suffers from an Oedipus complex, is not a text that requires the cooperation of an ideal reader to make its unconscious tendencies evident. The revelation of those tendencies has nothing to do with the process of textual cooperation. It has to do rather with a subsequent phase of approaching the text in which, after semantic actualization, it is evaluated and criticized. The criticism can evaluate its "esthetic" success (whatever definition is given of that effect) and the relations between the ideology and stylistic solutions of the author and economic situation, and can search for the unconscious structures manifested through the actantial structures (without their having been the content understood by the author). However, such psychological, psychiatric or psychoanalytic inquiry, though important and fruitful, would belong with the use of the text for documentary purposes, and is located in a phase subsequent to semantic actualization (even if the two processes can in turn affect each other). It is as if in the expression "I confess everything" there were a matter of textual cooperation to actualize all its semantic implications, to define its topic, and to single out its remote presuppositions. On the other hand there is also a matter of documentary utilization to use this expression as the judicial evidence that the speaker was guilty. But that would mean that in the expression "Come here, please" it is not a matter of textual cooperation to infer that the speaker is motivated by the evident desire to have me go to him. On the contrary, it appears that this type of inference is an essential part of the actualization of the message. We posit that there does exist a text in which the author could clearly not have been aware of encyclopedic data through which a whole series of actions or relations express given psychic contents, and yet it appears quite evident that the entire textual strategy leads inexorably to invest it with contents of that nature. A typical case could be Sophocles' Oedipus Rex, at least the way Freud read it. It is clear that we today can read the tragedy as related to an encyclopedia containing, across its own sub-

16 TWO PROBLEMS IN TEXTUAL INTERPRETATION 159 codes, the results of the Freudian overcoding, but it must be said that neither Sophocles as subject of the utterance nor Sophocles as textual strategy could relate to that encyclopedia. Nevertheless, the very blind obstinacy of Oedipus in dismissing the truth that is offered to him at various times in such an incontrovertible manner, seems the primary content of the Sophoclean text. Let us say that in this case the author was instituting new encyclopedic data. The text as an act of invention (see the definition of that category in Eco, 1976: ff) institutes a new code, and for the first time establishes a relationship between the expressive elements and the content data that the semantic system up to that point had not yet defined and organized. In that case, the Freudian reading constitutes a legitimate operation of textual cooperation, which actualizes what there is in the text and what the author as expressive strategy puts there. For after the empirical Sophocles as subject of the utterance was more or less aware of what he was doing textually, it is a matter of use, of symptomatological reading, and extraneous to the activity defined by a theory of textual cooperation; it applies, if we like, to Freud as Sophocles' personal physician and not to Freud as Model Reader of Oedipus Rex. And that leads us to say (or reiterate) that the Model Reader of Oedipus Rex was not the one Sophocles was thinking of, but the one Sophocles' text postulates. It is likewise clear that the Sophoclean text, in postulating the appropriate Model Reader as cooperative strategy, constructs a reader capable of bringing out those content data which up to then remained concealed (if one admits that Sophocles was the first to take note of the manifestations that are called the Oedipus complex, and the encyclopedia of Greek culture of the period did not yet have competences organized for the purpose, say intertextual traditions). In other words, the Model Reader of Oedipus Rex is required to carry out cooperatively the same operations of recognitions of relations that Oedipus as a character is asked to carry out - and that he does somewhat late. In that sense certain narrative texts in telling the story of a character meantime provide pragmaticsemantic instructions to their Model Reader, whose own story they tell. It is reasonable to suppose that to a certain extent this happens in any narrative text, and perhaps in many others that are not narrative. De tefabula narratur. To illustrate better the difference we are trying to isolate, let us take an example from the interpretation Maria Bonaparte (1949) gives to Edgar Allan Poe's work. She demonstrates that Poe (already defined by Lauvriere as an extreme degenerate and by Probst as an epileptic) was totally impotent. He was dominated by the impression he felt as a child when he saw his mother, dead of consumption, on the catafalque. Thus, in adult life he always felt morbidly attracted, in fiction as well as fact, to women with the unhealthy and funereal features of his mother. From this memory came his necrophilia, his love for diseased child-women, and his stories populated by living corpses. Naturally, the data is taken indiscriminately from the poet's life and from his texts, a procedure that is not incorrect for a psychological study of a person called Edgar Allan Poe, but is to be rejected for a study on the Model Author that the reader of those texts imagines and needs to imagine even if he has no

17 160 UMBERTO ECO biographical data on Edgar Allan Poe. In that case we can safely say that Maria Bonaparte is using Poe's texts as documents, symptoms, and psychiatric reports. It is a pity she could not do so with Poe alive, and thus help to cure him of his obsessions. Be that as it may, the fault is not hers, and as Poe is dead, we retain the very human and also scientifically productive satisfaction of reflecting on the exemplary case of one of the greats, and on the mysterious connections between disease and creativity. All that has nothing to do with a semiotics of the text or an analysis of what the reader can find in Poe. But Maria Bonaparte also knows how to do a textual semiotics, and splendidly. In the same essay, she analyzes the poem "Ulalume": the poet wishes to reach the planet of Venus-Astarte; earth-bound Psyche holds him back. He follows his path nonetheless, but at the end of the journey finds the tomb of his loved one. Maria Bonaparte observes that the poet's symbolism is very transparent, and makes a sort of actantial analysis ante litteram: a dead actor prevents Poe from approaching love that is psychically and physically normal, symbolized by Venus. Transform the actors into plain actantial polarities and we have a subject observing an object, a helper and an opponent. When Maria Bonaparte examines the tales "Morella," "Ligeia" and "Eleonora" she finds that all three have the same fabula (following in this respect some intuitions of Baudelaire). Aside from a few differences, we always have a husband in love with an exceptional woman. She dies of consumption, the husband swears eternal grief, does not keeep his promise, and becomes attached to another person; but the dead one reappears and wraps the new woman in the mantle of her funereal power. From that fabula (a true intertextual "script") it is easy to pass to actantial structures, and Maria Bonaparte does so instinctively when she considers the second woman of the third tale dead as well: though she does not die, in some manner she plays the role of love object who escapes from the lover, and thus identifies herself with the first woman. Maria Bonaparte perceives in the three tales the structure of an obsession, and perceives it first of all as a textual obsession. But this beautiful analysis is continuously interwoven with biographical remarks that connect the textual evidences with the various aspects of Poe's private life. So Maria Bonaparte takes a methodological detour that diverts her attention from the interpretation of the texts to their use on a clinical plane. The image of the mother, which does not appear in the texts, is introduced within their framework from extratextual sources. Let us now see reading of an opposite type more akin to our conception. It is the one done by Jacques Derrida on "The Purloined Letter" in Le facteur de la verite (1975), dealing with both the reading of Bonaparte's and the celebrated one by Lacan (which on the other hand he criticizes). Starting with an appropriate ideological competence that leads him to favor the discourse of the unconscious in the text, he identifies in it more general subjects of the actors that represent them. He does not count so much the nature of the letter as the fact that it gets back to the woman from whom it was taken, or that it was found hanging from a nail under the center of the fireplace ("on the immense body of the lady, between

18 TWO PROBLEMS IN TEXTUAL INTERPRETATION 161 the posts, between the legs of the fireplace"); and he does not count so much the actor Dupin as the fact that he displays a deceitful character by which "he identifies successively with all the characters." It is not a matter here of deciding whether Derrida's interpretation satisfies the plurality of possible contents exhibited by Poe's text. What is interesting here is that Derrida wishes to uncover, as he says (and in opposition to the position he imputes to Lacan) the "textual structures"; he does want to "question Poe's unconscious" but "not the author's intentions," and in order to do so he seeks to identify it from time to time "with this or that position of its characters." In doing so Derrida proceeds from the fabula (chosen according to his own ideological leanings, which guide him in individuating what for him is the topic of the whole event, a story of castration) to the actantial structures, showing how they are manifested at the deep levels of the text. Good or bad, the process is legitimate. It remains to be said whether this method of proceeding should not be viewed more as critical interpretation than interpretive cooperation. But the borderlines between these two activities are very narrow and are set in terms of cooperative intensity and of clarity and lucidity in expounding the results of an actuated cooperation. The critic in this case is a cooperating reader who, after actualizing the text, relates the appropriate cooperative steps and renders evident the way in which the author, through the proper textual strategy, led him to cooperate in that way. Or else, he evaluates in terms of esthetic success (however, theoretically, he defines it) the modality of the textual strategy. The methods of criticism are varied, we know: there is philological criticism, esthetic criticism, sociological criticism and psychoanalytical criticism; criticism that expresses value judgements and criticism that throws light on the direction of a piece of writing. And others besides. The difference that is of interest here is not that between textual cooperation and criticism, but between criticism that relates and yields the modalities of textual cooperation, and criticism that uses the text, as we have seen, for other purposes.